Constructing a Corpus

“..a stick thrust in water felt straight and looked bent to a Greek. The sun moved for the Inquisition, the earth for Galileo. Light is a wave for Schroedinger and a particle for Heisenberg. But even the last have had their Dirac. The seeming contradictions vanish in the grace of greater knowledge. We have learned that the answer depends upon how we ask the question. And we have learned to ask the question so as to get an answer of a kind that we can use….”
Warren S. McCulloch
Through the Den of the Metaphysician


The four central studies of this book are about Miriam, my daughter. The extensive detailed data collected in the study of her thinking and the attempt to use it all in interpretation represent an effort to tighten up the case study approach to permit empirically based descriptions of cognitive structures. The analyses of Miriam’s learning form a unified exploration of how local changes in cognitive structure result in significant large scale effects. The study was inspired in part by a suggestion of Flavell’s (1963) to blend Piaget’s focus on cognitive structures with the ecological emphasis of Barker and Wright and undertake “a type of research endeavor which has not yet been exploited: an ecological study of the young child’s mundane interchanges with his workaday world.”

Because this book is not about a great man’s thought but merely about the marvel of a normal child’s learning, it depends for its general interest in a technical way on arguments about the lawfulness of psychic phenomena advanced by the psychologist Kurt Lewin. He argued that the individual case does not merely illustrate the general law; it embodies the general law. If mental phenomena are lawful in a strong sense, as physical phenomena are, one can arrive at the general law through detailed interpretation of the particular case [note 1]. The primary objective of The Intimate Study was to produce a corpus which could be analyzed in such a way as to advance our understanding of learning. Let me first describe what that corpus is and how it was constructed then explain some of the reasons and accidents that gave the project its specific form.

Activities and Observations Recorded

At the beginning of The Intimate Study, we were four in my family, two parents and two children, Robby, age 8, and Miriam, age 6. The specific objective we followed was to trace in fine detail Miriam’s learning for the six months following her sixth birthday (April 9th through October 8th, 1977). This is the CORE of The Intimate Study; it is extended beyond this core period by later observations. I recorded her behavior in well structured situations and followed her beyond the confines of the computer laboratory with naturalistic observation in the various settings of her everyday world. Miriam was under continual observation for six months.

An Unschooled Child and Continually Available Subject

When The Intimate Study began, Miriam was attending kindergarten (one making no substantial academic demands upon her). As the study ended, Miriam completed her first month of first grade. Miriam was an unschooled school-age child. I frequently visited kindergarten and was a part of her social world as well. When her friends came to play, on rainy days there was no place to escape them in our small dwelling; on sunny days, they often asked me to join them at the tree fort, in the garden, playing in the courtyard, or pushing them on the space trolley. We lived in the carriage house of an old suburban mansion. By distance and dangerous roads, in effect, the children were imprisoned on the mansion’s spacious grounds unless they could get a parent to act as chauffeur. Very occasionally, the children would visit friends or go shopping with their mother. But over the summer, most of Miriam’s friends were away, and their mother preferred shopping alone.

With very few exceptions, the only times Miriam spent away from home were in my company; most of those times were the hours at project Logo’s Children’s Learning Lab. Her young age, the limits of her world, and my having the time permitted us to share a common world for the period of The Intimate Study. The data collected during this period are grouped into four matters: profiles; sessions; vignettes; and the log.

The Profiles

The profiles are a series of initial and terminal cognitive examinations. Through these, Miriam’s capabilities and styles of thought may be compared with data in the psychological literature and with normal education skills through her performance of these specific tasks [note 2]:
– School tasks: reading and arithmetic skills.
– Piagetian tasks: one-to-one correspondence; class inclusion; continuous quantity; time; object volume; combinations; substance; weight; displacement volume; backspinning a ball; beam balance; multiple seriation.
– Stanford-Binet Intelligence Scale: test L-M at 6 years, 19 days.
Other materials of an idiographic character were collected but are not reported here.

The Sessions

The sessions were mechanically recorded, all on audiotape and many on videotape. Processing of the marginally edited transcripts and working materials of these sessions is described below. These relatively formal, directed working sessions at MIT’s Logo lab (approximately 70) and at home (approximately 20) exhibit more than four months of interactions between Miriam and me in a computer-centered environment. The data are as detailed as any critic could wish.

The Vignettes

Richly interpretive, highly subjective, open to error and overstatement, these materials are essentially ephemeral literary constructs whose purposes are to document events in the social world of our family and to connect themes emergent in the more structured data. The vignettes are like snapshots of thinking or short stories that surfaced in the small society of our family. They are based on selective naturalistic observation of Miriam’s behavior beyond the range of mechanical recording and in situations where the recording itself would have been obtrusive. I attempted to capture all unrecordable and significant expressions of development Miriam exhibited during The Intimate Study.

To the extent that they record observations by an ever-present scientist in the midst of the action, they attempt to elevate anecdotal reportage to the status of naturalistic observation through the claim that they record ALL thematically interesting behavior in those settings beyond the range of mechanical recording. Though imperfect, these data may still be accepted as additional, well-placed pieces in the puzzle, pieces that have in fact been essential in helping me grasp patterns in the development of Miriam’s mind. Each vignette includes a short sketch of the point of view from which I judged the content significant.

The Log

These daily notes for the last five months of The Intimate Study record how Miriam spent her time. The observations are quite variable in level of detail and quality. The objective was to note what Miriam was doing every half hour or hour. Lacunae exist (occasions where I was asleep or otherwheres) and even a day or two may have been missed. No analysis of these observations has been attempted here.

Processing Observations of the Corpus

A description of our typical day during The Intimate Study will illuminate how the data were rendered in their first processed form. Early in the morning, the children and I would ride to the Logo Lab and execute our day’s experiment. Each session was mechanically recorded; printed output was collected and labelled with session numbers. We would drive home for lunch, following which the children would occupy themselves with their own amusements for the remainder of the day. I would spend several hours transcribing the recordings of Miriam’s work in manuscript. After that, I composed vignettes, attempted to understand ongoing work, and planned future sessions. Usually this work occupied me until late in the evening.

Throughout the day and outside the laboratory, I would interrupt this processing to note in the Log what Miriam was doing. The essential value of the Log was in returning me regularly to observe Miriam’s activity. The lag time to transcription for recordings was usually kept down to several days at the maximum. For vignettes, the lag was more variable but usually under a week. Those days when the children did not want to work, I was able to catch up. Consequently, the data were processed to this rudimentary state in nearly real-time. Figure I summarizes the elements of the corpus and how the materials relate to the analyses in this book.

Contrasting the Study of Learning with the Evaluation of Knowledge

The study of the process of learning is related to but different from the assessment of states of knowledge. I claim that The Intimate Study represents a sensible approach to studying the processes of learning, despite the manifest difficulty of the task and the method’s vulnerabilities to criticism. Further, I claim there is no other empirical approach with such promise of telling us anything important about the accommodation of mental structures through experience in the language-capable mind. If the experimental task is too simple, observations may lead erroneously to a simple-minded view of mind. Even if the mind should prove to be a simple thing (which I doubt) its interactions with experience may be much more complex than studies in the Peterson paradigm or of cryptarithmetic would suggest (the former involves memorizing lists of nonsense syllables, the latter decoding puzzles where letters have number values).

Some Reasons for Adopting the Method

There are at least three fundamental problems in exploring learning as insight. Major insights happen rarely; they can not be scheduled; and they may not be recognized when they do occur. All three problems set major limits on any method which could possibly be useful for studying the accommodation of cognitive structures . During the nine months wherein I followed Miriam’s learning closely, she learned how to add. For her, this learning required four insights (one about every ten weeks!). This witnesses the rarity of such learning. Contrasting the small number of insights with the hundreds of pages of transcriptions which capture her calculation behavior exemplifies the ‘noisy data’ problem.

An insight is something that happens in an individual mind; it is a private experience. There may be manifestations in behavior, e.g., a shocked look or a surprising question, but the best evidence available is the subject’s witness as to the occurrence and significance of the event. Consider this example:
Where I was reading, Robby assembled a puzzle on the living room floor. He left off the puzzle and lay on his side, bending his body back and forth at the pelvis. When I asked him to stop squirming, he sat up and spoke to me:

Rob Daddy? You know all that stuff about 3 hundred and 60 ? (A reference to discussions of the effect of reducing an angle by 360 degrees.) I understand it now.
Bob Wow! How did you figure it out?
Rob Well, you know if you have an angle that’s 3 hundred and 61?… And you take away 360? It’s 1, and that’s like it’s starting all over again.
Bob That’s really great, Rob. When did you figure it out?
Rob Now.
Bob Just now? When you were squirming around there on the floor?
Rob Yeah. Squirming around helps me think.

Robby returned to his puzzle.

If Robby had not been willing and eager to reveal his thoughts, his understanding of this particular problem at this time would have been missed; more importantly, the evidence of the connectedness of his knowledge would have been missed as well. Had it not surfaced in our conversation, who would have guessed that his squirming around on the floor in a place and at a time remote from our formal sessions was a response to an earlier confusion based on work at Logo? I would not have realized his squirming was a simulation, with his own body, of operations performed by the turtle in experiments at the lab. Surely such symbolic behavior as this ‘simulation’ is evidence about how knowledge in one structure relates to another.

Principles the Study Embodies

The preceding ideas and examples suggest why I adopted the methodology of The Intimate Study. Let me state as lucidly as possible the principles that study embodies:

  1. The most significant form of learning is the accommodation [note 3] of cognitive structures; underthe assumption that insight is the phenomenological correlate of this accommodation, the functional objective of The Intimate Study is to explore the interaction of mental structure and experience when learning-as-insight occurs.
  2. Because insight occurs rarely, the study must be long to assemble a sufficient collection of incidents of learning for analysis.
  3. Because insight can not be scheduled, the observations of the subject must be continual and protracted; the data collection must be an everyday part of the subject’s life in which he participates willingly for his own reasons.
  4. Because insight is essentially a private event, the subject must trust the experimenter and be a colleague more than an object; reports from articulate introspection are a key source of information.
  5. To the extent that insight involves the integration of ideas, it requires a rich interpretation of a sort typically not accessible to the subject; while the subject experiences it as personally significant, the experimenter will view it as theoretically significant.
  6. Inasmuch as the experimenter has an imperfect theory of mind, is insensitive to the importance of specific incidents, or can not comprehend the mass of observation as it is developing (these conditions are always true), the strategy of choice is to create a corpus of sufficient richness and permanence that it may be queried as subsequentinterpretation proceeds.
  7. The final point is primarily an ethical one. One should not establish a relation of trust in a prolonged and intimate study of cognitive development and then abandon the subject when the experiment is finished. Respect for the subject, even love, is appropriate; one must not take advantage of the subject [note 4].

On Limited Objectivity

Some critics will feel that the study is reduced in value by a lack of scientific objectivity. I don’t feel that one should be overly concerned with objectivity, especially because it would prove impossible for a study at this level of detail [note 5]. Many of the ideas and objectives which guided the execution of The Intimate Study are novel and sound, but some are certainly at least partly wrong. So long as errors are not disguised, the harm in them is merely two-fold. First, the value of the corpus, compared to a perfectly executed study, will be diluted, and thus the conclusions will be less compelling than they might otherwise have been. Second, I will appear more stupid than I like to admit being [note 6].

Accidents Affecting The Intimate Study

Some of the accidents involved in the genesis of The Intimate Study are the following. For two years, Robby had come often to the Logo Laboratory and worked with me there, giving me his reactions to programs I made and ideas I wanted to try out. Those projects, thematically various and undertaken with no overall organization in mind, comprise a pilot to The Intimate Study in fact even though not in intention. That earlier work had impressed me with the importance of long studies, the usefulness of mechanical recording, and the value of an approach which could capture unscheduled learning. This latter idea is central to The Intimate Study: the accommodation of cognitive structures happens rarely but is the most significant form of learning. With this point of view forward in my mind, I stumbled upon Flavell’s suggestion in a discussion of the stability of Piagetian stages, which I cite here in extension for its importance in forming the objectives and methods of The Intimate Study:

“The question which these investigations bequeath to us is the same one they were designed to answer: how do conservation and other Piagetian concepts develop? It might be said, parenthetically, that this question can in principle be divided into two: by what concrete methods deriving from what theory of acquisition can we most effectively train children on these concepts? and — what may be a different question — how are they in fact acquired normally — that is in the child’s day-to-day cognitive bouts with the real world? It is not likely that the second question can ever receive a direct and precise answer because of the intrinsic difficulty of sorting out the myriad uncontrolled environmental presses to which the child is exposed. But the very posing of the question in this way suggests a type of endeavor which has not yet been exploited: an ecological study of the young child’s mundane interchanges with his workaday world, a kind of job analysis of his daily life along the lines of Barker and Wright’s work (1951, 1955) [note 7]. This sort of study might suggest promising leads to pursue in the more conventional transfer of learning experiment; in view of the infertility of most training methods so far tried, no source of new ideas should be ignored.”
John Flavell
The Developmental Psychology of Jean Piaget

The Intimate Study was an attempt to approach a variant of that impossible second question: what is the process of learning like in the language-capable mind. The study shared with ecological psychology the objective of specifying how components of mind relate to the specifics of experience. Such a challenge I considered an opportunity of the first order and one congenial to my interests and values, and to my obligations as well.

Miriam and Rob: Colleagues more than Subjects

During the period of the study, my wife and I planned to have another child (now our daughter Peggy), and I expected the responsibility for the older children would be mine. How much better an idea if we three should undertake together a significant research project than that we should merely entertain or tolerate each other for the duration. The Intimate Study was thus an experiment ‘with’ children, not an experiment ‘on’ children. It was not undertaken to make them academic prodigies. Miriam and I shared a robust love and respect. Her brother occupied no less a place in my heart. They were willing to help me with my work because doing so implied we would spend a lot of time together. In function, we three comrades worked like a project team or task force. If I was first among equals because of my specific skills and depth of experience, nonetheless Robby and Miriam exercised considerable direction in our daily doings because my objective was to follow in detail their understanding of our work. Some work bored them; some engaged and inspired them to products they are still proud of. I do not believe the work has damaged them in any way. On the contrary, our involvement in the project certainly enriched their lives. Such is my personal and ultimate conclusion.


It is a commonplace of relativism that every system of ideas is at root a metaphor from which its vision grows, branches, and is nourished. In such a view, the appropriate criterion for the value of a system of ideas is fecundity in application to cases we care about. Many people assume that computational descriptions offer no substantial advance to psychology but are merely the latest fad in the reformulation of age old problems. On the contrary, with the founding of cybernetics Norbert Wiener and Warren McCulloch began a ‘Copernican’ revolution in the study of mind, one crisply phrased in this assertion: the laws of the embodiment of mind should be sought among those governing information rather than those governing energy or matter [note 8]. McCulloch, a neurophysiologist and logician, advanced an ‘existence proof’ for this position,

…a true mathematical idea: between the class of trivial combinatorial functions computable by simple Boolean logic and the too general class of functions computable by Turing machines, there are intermediate classes of computability determined by the most universal and natural mathematical feature of the net — its finiteness. This is pure mathematics. The theoretical assertion of the paper is that the behavior of any brain must be characterized by the computation of functions of one of these classes….
Papert, 1965
Introduction to Embodiments of Mind

This means less that ‘the brain is a computer’ than that it is possible in principle to create an adequate computational model of any brain. Notice that any model which described the organization and flow of information in a brain would describe identically what most people commonly mean by mind.

Experimental Epistemology

Near the end of his career, McCulloch spoke of turning psychology into experimental epistemology by attempting to understand the embodiment of mind. This task he understood as involving three problems:
a physiological problem (understanding the brain),
a psychological problem (understanding the mind),
and a logical problem (understanding how to represent relations such as exist in both mind and brain).
He required of this logic that it be able to represent both the organization of mind, in detail, and possible different states of that organization. His example: descriptions of cognitive structures should be able to represent the mental equivalent of the fist as well as the hand. The step-wise trace of a program’s execution (they are easy to acquire) bears such a relation to the text of the program; it represents one possible performance of a functioning program. It is this aspect of computational descriptions — that an execution trace can represent functioning in a particular case — that permits addressing specific problems, e.g. how a particular mind was changed by a specific experience [note 9]. Through the addition of functioning as an essential aspect of structure, representing the mind as programming structures provides a specific way to make sense of Piaget’s proposal that structure and development are correlatively generative; that is, new structure develops through the functioning of pre-existing structure.

The Elevation of Control and the Relational Conversion

Papert has proposed that the insertion of control elements into functioning structures is the “missing link” in descriptions of knowledge needed to understand significant learning. I ask if one can apply the essentially mathematical ideas of McCulloch and Papert to solve particular problems of learning [note 10]. Could one use a variant of Papert’s idea, which I call ‘the elevation of control,’ to explain in detail an individual’s mind as a system evolved through the experience of that individual interacting with his everyday world? [note 11]

Nothing comes from nothing; what are the circumstances through which minimal means can engender connections between preexisting structures? Going beyond Papert’s proposal, I introduce in the following analyses another idea which can help explain the creation of new links and structures in the mind. Call it the ‘relational conversion.’ Imagine there exist within the mind large scale cognitive structures competing with one another to solve problems. Each attempts to select aspects of the problem as presented to the mind on which its ‘programs’ can work. Assume that in order to prevent paralyzing confusion, when one structure is able to appropriate a problem it is also able to inhibit competitors, as if by sending a message ‘you must not confound this situation (which is my problem) with other situations.’ Such a scenario is the basis for the establishment of ‘must-not-confound’ links between disparate but competing cognitive structures. Learning, seen as significant structural change, occurs when one of these ‘inhibitory’ links is converted into a ‘relational’ link through an insight contingent upon a specific experience. This begins the coordination of disparate cognitive structures, a theme of interpretation throughout this book. It is also a simple and direct explanation of how new structure can be created almost from nothing.


A very simple model of how to do science proposes that hypothesis confirmation and disconfirmation is the central activity. One might then contrast two grand theories, draw inferences from them applicable to a single case, then test the implications of the two different theories as to which the evidence makes appear the more worthy of belief. A richer perspective is presented by Goodman wherein one recognizes that a primary factor which separates the true from the logical is the choice of categories through which things are described. This has precious little to do with any sort of valid inference, per se, and must be justified as right by some other means [note 12].

The distinction between valid induction and deduction, on the one hand, and how one divides the world with descriptions of things, on the other, does not imply that such choices about cutting up the world are theory-free or that studies which indulge in extensive exemplification are empirical in any mindless sense. How one divides the world with descriptions is the major decision from which theories and experimentation follow. But we can’t get at the content of the world, the thing itself; we can only ask what are the natural chunks of reality, how can we analyze or cut up the world for our better understanding in such a way that pieces separate individually without disintegrating. The criterion for judging descriptions is productivity: does the cutting produce a useful result ?

Rightness of Descriptions and Exemplification

The rightness of my descriptions of knowledge in the mind and its development is a central issue. In the detailed studies, I have resorted to extensive exemplification of the cognitive context to justify my arguments about the character of knowledge in the mind and the processes of its development. The major issue in exemplification is whether or not one’s examples comprise a fair sample — no matter from what material they are cut out [note 13]. What precisely are those things which the data-rich studies of this book exemplify ? By the chapters, they are:

Chapter The material of the study shows:
that non-expert problem solving is more like ‘bricolage’ than analysis.
how generally applicable skills can be understood as the controlled invocation of highly particular, domain specific knowledge.
that formal thought, in the Piagetian sense, is a competitor with concrete thought, not an emergent from its perfection.
that social engagement can engender intermediate (metastable) structures with functional lability from which develop in turn mature structures with their own more powerfully adapted stability.
how essentially different representations are brought into registration with one another, a requirement for any developing coherence of mind.

By choosing the exemplifying material with a focus on such themes, I am asserting that these are primary issues in understanding human learning.

Please email any comments directly to Thank you.

Text Notes:

  1. See the citation: “Lewin: On the Pure Case.”
  2. Summary data from these experiments are presented in appendix 1: “Relating this subject to Other Studies.”
  3. This is Piaget’s term for the process by which cognitive structures adapt themselves to new experiences. I use it in the same sense and try to describe it, partly through exemplification.
  4. When it began, this study was severely criticized with the argument that it might endanger the weel-being of my child; some feared she would grow up feeling used by a father whom she loved. There has been no such outcome in this case. To fail to consider such a possibility, however, would be irresponsible in the extreme. No one should undertake such a study withoput the firm, personal commitment that the well-being of the subject is more important than any outcome or even the continuation of the study.
  5. The arguments advanced in the citation “Langer: On Objectivity” may help make the minority position seem not unreasonable.
  6. I find some consolation in my favorite line by Anatole France from La Rotisserie de la Reine Pedauqe: “…je ne me flatte pas de direr grande honneur des ces revelations. Les uns diront que j’ai tout invente et que ce n’est pas la vrai doctrine; les autres, que n’ai dit que ce que tout le monde savait….”
  7. One Boy’s Day; Midwest and Its Children.
  8. For an appreciation of their contribution, see the article by Stephen Toulmin in The Horizon Book of Makers of Modern Thought.
  9. I do not claim to solve the problem of embodiment of mind, but I do hope to make some advance on the psychological front by applying some AI ideas, which comprise in ageneral a kind of logic of functioning representations, to approach Minsky’s goal of a theory of the emergence of control structure. I do not touch the issue of physiology; McCulloch’s colleague Lettvin now argues convincingly that the complexity of even a single neuron is still beyond the reach of AI and that to claim any reality for models of the ensemble is very rash and unwarranted.
  10. This book does not assume a knowledge of programming. I do refer to some ideas for representing aspects of thought which were almost inconceivable before the advent of computers, and I do apply computational ideas in a description of thinking and learning based on a s significant corpus. But my work is never formal for the sake fo formality, nor do I Produce programmed models of mind. My inferences from the data of my studies are too fragmentary to permit creation of functioning programs with excessive fabrication. To do so would be more deceptive than illuminating.
  11. The idea is more subtle than the mere insertion of links in a semantic net. The key point is the joining together with minimal means of two structures which own an essential epistemological complementarity. For Papert’s application of this idea to the analysis of Piaget’s conservation phenomena and its relation to Piagetian “groupements”, see the citation “Papert: On the Elevation of Control.”
  12. See the extended citation, “Goodman: On the RIghtness of Rendering.” The essence of Goodman’s position on e4xemplification is that it comprises one of the three primary modes of meaningful references (the others are representation and expression). The selection implicit n a thing’s being a sample determines what are the aspects in virtue of which it exemplifies and is thus meaningful.
  13. In Goodman’s Ways of Worldmaking, see Chapter 4 (When is Art ?). Here see also the extended citation “Goodman: On Multiple Worlds.”
Print Friendly, PDF & Email